Fifteen Canadian brands, read the way an AI agent reads them. One score out of 100 for each, built from what the server actually sends when an agent asks a customer’s question. Pages were read on 30 July 2026 from the server response, without a browser session.
The Canadian field separates on one decision: whether the rate is published into the page or injected by the browser. Three brands reach the leading band, ten sit in the building band, and two land early. Almost every gap measured is a rendering choice rather than a content shortfall.
Canadian financial institutions welcome AI agents. Fourteen of the fifteen brands measured allow every named AI crawler in their robots file, and thirteen serve those agents the identical page a browser receives. Permission is not the constraint in this market.
The constraint is what arrives. Across the field, the pattern repeats with unusual consistency: the page structure ships correctly, the table headers ship correctly, the row labels ship correctly, and the number is waiting on the browser. RBC serves 91 rate spans ready for values. TD serves 48. CIBC serves 45 template tokens across a fully built mortgage table. Tangerine serves square brackets.
Three brands break the pattern in the other direction. Wealthsimple publishes every price, tier threshold and FAQ answer as served text and scores 88. RBC and TD both clear 70 on the strength of card, deposit and insurance pages that answer completely. Two brands sit at the base of the field for the same reason in reverse: American Express Canada and Capital One Canada run open robots files and clean edges, then serve card pages that reduce to navigation. Both were verified against a second independent retrieval pass before publication.
The commercial point is straightforward. An agent asked to compare Canadian mortgage rates today can quote three brands. Whichever brands publish those numbers first become the reference answer agents repeat, and reference answers are slow to change once set. Nothing here requires a rebuild, since the content is written and the tables are already built.
Each brand was scored on the surfaces it actually sells, up to seven. The institution score is the average across those surfaces. Select any brand to open its full deep dive.
| Rank | Institution | Category | Tech /50 | Content /50 | Surf. | Score | Band |
|---|---|---|---|---|---|---|---|
| 1 | Wealthsimple | Fintech | 41 | 47 | 7 | 88 | Leading |
| 2 | RBC Royal Bank | Big Five bank | 34 | 38 | 7 | 72 | Leading |
| 3 | TD Canada Trust | Big Five bank | 42 | 28 | 7 | 70 | Leading |
| 4 | Tangerine | Direct bank | 40 | 28 | 7 | 68 | Building |
| 5 | Fairstone Financial | Consumer lender | 35 | 31 | 7 | 66 | Building |
| 6 | Scotiabank | Big Five bank | 36 | 30 | 7 | 66 | Building |
| 7 | Desjardins | Cooperative | 31 | 30 | 7 | 61 | Building |
| 8 | BMO | Big Five bank | 29 | 31 | 7 | 60 | Building |
| 9 | Questrade | Brokerage | 31 | 28 | 7 | 59 | Building |
| 10 | CIBC | Big Five bank | 31 | 27 | 7 | 58 | Building |
| 11 | EQ Bank | Direct bank | 27 | 31 | 6 | 58 | Building |
| 12 | National Bank of Canada | Big Five bank | 33 | 25 | 6 | 58 | Building |
| 13 | Neo Financial | Fintech | 29 | 24 | 7 | 53 | Building |
| 14 | American Express Canada | Card issuer | 16 | 9 | 7 | 25 | Early |
| 15 | Capital One Canada | Card issuer | 17 | 7 | 7 | 24 | Early |
Brands compete against the set a customer actually shortlists against, so ranking within category is where the commercial conversation sits.
Wealthsimple, RBC and TD Canada Trust answer the customer question in full on most of their surfaces.
Wealthsimple, Neo and EQ Bank serve a full mortgage rate table an agent can read. Desjardins and Tangerine add hero rates.
TD, Tangerine, Fairstone and American Express publish a machine-readable index. The TD file covers 436 Canadian pages with real descriptions.
Each finding below rests on evidence captured from the served response of the brands named, read on 30 July 2026 without a browser session.
Eleven of fifteen brands ship a fully built rate table where the structure, the term labels and the APR columns all arrive correctly. Filling those cells server side is the highest value change available to almost every brand measured, and it moves several surfaces at once.
Annual fees, monthly fees, purchase rates and qualifying income publish as readable text at RBC, TD, Scotiabank, Tangerine, National Bank and BMO. The publishing pattern that gets a card page right already exists inside each of these organizations and transfers directly to the rate pages.
Wealthsimple and Fairstone serve FAQ answers as text with schema on nearly every surface. Neo carries 61 questions ready for the answers that already sit in its client payload, and National Bank holds its answer text in accordion targets ready to render. FAQ answers are the exact content agents quote, which makes this the cheapest content gain in the field.
EQ Bank publishes an open robots file and serves Googlebot normally, so extending the same treatment to named AI crawlers matches the server to the stated policy. American Express Canada serves a consent gate where card details sit, so serving product facts unconditionally puts fee and rate in the first response. In both cases the written content is strong and the delivery decision is what the score reflects.
The reference answer an agent gives about Canadian banking in 2027 is being written by the pages published this year.
Each deep dive carries the score hero, the short version, the read, the surface scorecard, what is working, where the value sits, field position, the 90-day path and the methodology.
AI Compass measures what an AI agent receives when it visits a financial institution website on a customer’s behalf. It reads the served response the way an agent does, without a browser session, without scripts, and without a human clicking to reveal a value.
Each product surface earns a score out of 100, built from Technical Readiness out of 50 and Content Architecture out of 50. The institution score is the average across the surfaces reviewed, up to seven.
Sub-categories come from the personal banking navigation on each brand’s Canadian root. Within each one, the flagship is the product the institution itself features first. Brands with fewer product lines are scored on the surfaces they actually sell, so a focused brand is measured on quality rather than on breadth.
Pages were read on 30 July 2026. Rates, bonus offers and terms move, and scores describe what each site served on that date. Where a site offers more than one language, the version served by default to a request with no language preference was used.
Scores come from observed page behaviour alone. No institution reviews or approves a score before publication, and every brand is measured against the same rubric. Outlier results are verified against a second independent retrieval pass before publication.
The score covers the public marketing site. Logged-in banking, the mobile app and the branch experience sit outside it. A site can score well here and still have work to do inside the login, and the reverse holds too.